Harry in a hurry? Roque claims ‘victory’ based on incomplete COVID-19 data
University of the Philippines professors who belonged to the OCTA Research group of experts had estimated that the cases could shoot up to 40,000 by June 30.
The researchers – among them, mathematician Guido David, political scientist Ranjit Singh Rye, and Maria Patricia Agbulos – had a track record on these projections, having studied the data of the Department of Health (DOH) since the outbreak began. In early May, they projected: 24,000 cases by June 15. DOH data showed 25,392 confirmed cases by June 13.
So when, in the morning of June 30, the DOH figure for confirmed cases was at 36,438 (from June 29), Roque declared, his fist in the air: "Wala na po, panalo na tayo. We beat the UP prediction po. We beat it! So, congratulations, Philippines!"
Later that day, the DOH reported a total of 37,514 confirmed cases.
Not real-time situation
What the presidential spokesman didn't acknowledge was that as early as June 28, the tally of the DOH showed over 46,000 positive cases. In the June 30 afternoon briefing, the health agency said there were 37,514 confirmed cases.
The daily tallies of the DOH distinguish between confirmed cases and positive cases.
The positive cases are positive test results reported by laboratories but have yet to be validated by the DOH. These positive cases are only added to confirmed cases once they are validated by the agency.
Due to backlogs in testing and validation – an issue the DOH also acknowledges – the confirmed figures they release therefore don't account for the real-time situation of the coronavirus outbreak in the Philippines.
How are backlogs a factor?
The DOH has been working for over a month now to clear its backlog of unvalidated tests. During this validation process, the department checks positive test results reported by labs for possible duplicate cases or other encoding errors.
The DOH earlier noted the difference between positive and confirmed doesn't necessarily reflect backlog of confirmed cases as these may contain duplicate cases.
Aside from validation backlogs, there are also testing backlogs, which include samples that have already been collected by labs but have not yet been processed. In this scenario, a person whose samples have already been collected can either be negative or positive for the virus.
UP researchers estimate that at least 15% of confirmed cases will account for the positive number of cases from the testing backlog. This rate is based on computations made from past data from the start of the outbreak up until June 19.
“The implications of this in estimating real time data is to use case reports with a loading factor to account for the testing lag, although this will not account for the DOH backlog,” the researchers said.
These estimated positive cases will then be added to the number of confirmed cases once they are processed and validated.
How estimates are computed
According to David, this means that, if June 30’s tally of confirmed cases is 37,514, then we can expect at least 15% of 37,500, or at least 5,625 cases, will be added to the tally once samples collected are tested.
The total cases as of June 30 could therefore be at least 43,139. The exact number of real time cases will be known once the DOH completes the validation of its backlog.
The number of late cases or results of tests given to patients at least 4 days ago registered at 404 on July 1, 242 on July 2, 843 on July 3, and 1,091 on July 4, bringing the latest total number of confirmed cases as of June 30 to 40,094.
As of Sunday, July 5, confirmed coronavirus cases in the Philippines surged past 44,000, after a record 2,434 cases were newly reported.
In the reseachers' latest report, they project 60,000 to 70,000 cases by July 31 if the government doesn't come up with the appropriate strategy and intervention in response to the pandemic.
They said their model for projecting cases, which relies on estimates from DOH data, has been consistent with a less than 10% error.
"Despite its limitations, the model has a lot of value and has consistently been accurate in its forecasting capabilities, with an error less than 10%. Using a mathematical model is superior to having no model," the researchers said. – Rappler.com